1 // SPDX-License-Identifier: GPL-2.0-or-later <<
2 /* 1 /*
3 * lib/textsearch.c Generic text search in 2 * lib/textsearch.c Generic text search interface
4 * 3 *
>> 4 * This program is free software; you can redistribute it and/or
>> 5 * modify it under the terms of the GNU General Public License
>> 6 * as published by the Free Software Foundation; either version
>> 7 * 2 of the License, or (at your option) any later version.
>> 8 *
5 * Authors: Thomas Graf <tgraf@suug.ch> 9 * Authors: Thomas Graf <tgraf@suug.ch>
6 * Pablo Neira Ayuso <pablo@netfi 10 * Pablo Neira Ayuso <pablo@netfilter.org>
7 * 11 *
8 * =========================================== 12 * ==========================================================================
9 */ !! 13 *
10 <<
11 /** <<
12 * DOC: ts_intro <<
13 * INTRODUCTION 14 * INTRODUCTION
14 * 15 *
15 * The textsearch infrastructure provides te 16 * The textsearch infrastructure provides text searching facilities for
16 * both linear and non-linear data. Individu 17 * both linear and non-linear data. Individual search algorithms are
17 * implemented in modules and chosen by the 18 * implemented in modules and chosen by the user.
18 * 19 *
19 * ARCHITECTURE 20 * ARCHITECTURE
20 * 21 *
21 * .. code-block:: none !! 22 * User
22 * <<
23 * User <<
24 * +----------------+ 23 * +----------------+
25 * | finish()|<--------------(6)--- 24 * | finish()|<--------------(6)-----------------+
26 * |get_next_block()|<--------------(5)--- 25 * |get_next_block()|<--------------(5)---------------+ |
27 * | | 26 * | | Algorithm | |
28 * | | + 27 * | | +------------------------------+
29 * | | | 28 * | | | init() find() destroy() |
30 * | | + 29 * | | +------------------------------+
31 * | | Core API 30 * | | Core API ^ ^ ^
32 * | | +-------------- 31 * | | +---------------+ (2) (4) (8)
33 * | (1)|----->| prepare() 32 * | (1)|----->| prepare() |---+ | |
34 * | (3)|----->| find()/next() 33 * | (3)|----->| find()/next() |-----------+ |
35 * | (7)|----->| destroy() 34 * | (7)|----->| destroy() |----------------------+
36 * +----------------+ +-------------- 35 * +----------------+ +---------------+
37 * !! 36 *
38 * (1) User configures a search by calling t !! 37 * (1) User configures a search by calling _prepare() specifying the
39 * the search parameters such as the pat !! 38 * search parameters such as the pattern and algorithm name.
40 * (2) Core requests the algorithm to alloca 39 * (2) Core requests the algorithm to allocate and initialize a search
41 * configuration according to the specif 40 * configuration according to the specified parameters.
42 * (3) User starts the search(es) by calling !! 41 * (3) User starts the search(es) by calling _find() or _next() to
43 * textsearch_next() to fetch subsequent !! 42 * fetch subsequent occurrences. A state variable is provided
44 * is provided to the algorithm to store !! 43 * to the algorithm to store persistent variables.
45 * (4) Core eventually resets the search off 44 * (4) Core eventually resets the search offset and forwards the find()
46 * request to the algorithm. 45 * request to the algorithm.
47 * (5) Algorithm calls get_next_block() prov 46 * (5) Algorithm calls get_next_block() provided by the user continuously
48 * to fetch the data to be searched in b 47 * to fetch the data to be searched in block by block.
49 * (6) Algorithm invokes finish() after the 48 * (6) Algorithm invokes finish() after the last call to get_next_block
50 * to clean up any leftovers from get_ne 49 * to clean up any leftovers from get_next_block. (Optional)
51 * (7) User destroys the configuration by ca !! 50 * (7) User destroys the configuration by calling _destroy().
52 * (8) Core notifies the algorithm to destro 51 * (8) Core notifies the algorithm to destroy algorithm specific
53 * allocations. (Optional) 52 * allocations. (Optional)
54 * 53 *
55 * USAGE 54 * USAGE
56 * 55 *
57 * Before a search can be performed, a confi 56 * Before a search can be performed, a configuration must be created
58 * by calling textsearch_prepare() specifyin 57 * by calling textsearch_prepare() specifying the searching algorithm,
59 * the pattern to look for and flags. As a f 58 * the pattern to look for and flags. As a flag, you can set TS_IGNORECASE
60 * to perform case insensitive matching. But 59 * to perform case insensitive matching. But it might slow down
61 * performance of algorithm, so you should u 60 * performance of algorithm, so you should use it at own your risk.
62 * The returned configuration may then be us 61 * The returned configuration may then be used for an arbitrary
63 * amount of times and even in parallel as l 62 * amount of times and even in parallel as long as a separate struct
64 * ts_state variable is provided to every in 63 * ts_state variable is provided to every instance.
65 * 64 *
66 * The actual search is performed by either !! 65 * The actual search is performed by either calling textsearch_find_-
67 * textsearch_find_continuous() for linear d !! 66 * continuous() for linear data or by providing an own get_next_block()
68 * an own get_next_block() implementation an !! 67 * implementation and calling textsearch_find(). Both functions return
69 * calling textsearch_find(). Both functions <<
70 * the position of the first occurrence of t 68 * the position of the first occurrence of the pattern or UINT_MAX if
71 * no match was found. Subsequent occurrence 69 * no match was found. Subsequent occurrences can be found by calling
72 * textsearch_next() regardless of the linea 70 * textsearch_next() regardless of the linearity of the data.
73 * 71 *
74 * Once you're done using a configuration it 72 * Once you're done using a configuration it must be given back via
75 * textsearch_destroy. 73 * textsearch_destroy.
76 * 74 *
77 * EXAMPLE:: !! 75 * EXAMPLE
78 * 76 *
79 * int pos; 77 * int pos;
80 * struct ts_config *conf; 78 * struct ts_config *conf;
81 * struct ts_state state; 79 * struct ts_state state;
82 * const char *pattern = "chicken"; 80 * const char *pattern = "chicken";
83 * const char *example = "We dance the funky 81 * const char *example = "We dance the funky chicken";
84 * 82 *
85 * conf = textsearch_prepare("kmp", pattern, 83 * conf = textsearch_prepare("kmp", pattern, strlen(pattern),
86 * GFP_KERNEL, TS_ 84 * GFP_KERNEL, TS_AUTOLOAD);
87 * if (IS_ERR(conf)) { 85 * if (IS_ERR(conf)) {
88 * err = PTR_ERR(conf); 86 * err = PTR_ERR(conf);
89 * goto errout; 87 * goto errout;
90 * } 88 * }
91 * 89 *
92 * pos = textsearch_find_continuous(conf, &s 90 * pos = textsearch_find_continuous(conf, &state, example, strlen(example));
93 * if (pos != UINT_MAX) 91 * if (pos != UINT_MAX)
94 * panic("Oh my god, dancing chickens at 92 * panic("Oh my god, dancing chickens at %d\n", pos);
95 * 93 *
96 * textsearch_destroy(conf); 94 * textsearch_destroy(conf);
>> 95 * ==========================================================================
97 */ 96 */
98 /* =========================================== <<
99 97
100 #include <linux/module.h> 98 #include <linux/module.h>
101 #include <linux/types.h> 99 #include <linux/types.h>
102 #include <linux/string.h> 100 #include <linux/string.h>
103 #include <linux/init.h> 101 #include <linux/init.h>
104 #include <linux/rculist.h> 102 #include <linux/rculist.h>
105 #include <linux/rcupdate.h> 103 #include <linux/rcupdate.h>
106 #include <linux/err.h> 104 #include <linux/err.h>
107 #include <linux/textsearch.h> 105 #include <linux/textsearch.h>
108 #include <linux/slab.h> 106 #include <linux/slab.h>
109 107
110 static LIST_HEAD(ts_ops); 108 static LIST_HEAD(ts_ops);
111 static DEFINE_SPINLOCK(ts_mod_lock); 109 static DEFINE_SPINLOCK(ts_mod_lock);
112 110
113 static inline struct ts_ops *lookup_ts_algo(co 111 static inline struct ts_ops *lookup_ts_algo(const char *name)
114 { 112 {
115 struct ts_ops *o; 113 struct ts_ops *o;
116 114
117 rcu_read_lock(); 115 rcu_read_lock();
118 list_for_each_entry_rcu(o, &ts_ops, li 116 list_for_each_entry_rcu(o, &ts_ops, list) {
119 if (!strcmp(name, o->name)) { 117 if (!strcmp(name, o->name)) {
120 if (!try_module_get(o- 118 if (!try_module_get(o->owner))
121 o = NULL; 119 o = NULL;
122 rcu_read_unlock(); 120 rcu_read_unlock();
123 return o; 121 return o;
124 } 122 }
125 } 123 }
126 rcu_read_unlock(); 124 rcu_read_unlock();
127 125
128 return NULL; 126 return NULL;
129 } 127 }
130 128
131 /** 129 /**
132 * textsearch_register - register a textsearch 130 * textsearch_register - register a textsearch module
133 * @ops: operations lookup table 131 * @ops: operations lookup table
134 * 132 *
135 * This function must be called by textsearch 133 * This function must be called by textsearch modules to announce
136 * their presence. The specified &@ops must ha 134 * their presence. The specified &@ops must have %name set to a
137 * unique identifier and the callbacks find(), 135 * unique identifier and the callbacks find(), init(), get_pattern(),
138 * and get_pattern_len() must be implemented. 136 * and get_pattern_len() must be implemented.
139 * 137 *
140 * Returns 0 or -EEXISTS if another module has 138 * Returns 0 or -EEXISTS if another module has already registered
141 * with same name. 139 * with same name.
142 */ 140 */
143 int textsearch_register(struct ts_ops *ops) 141 int textsearch_register(struct ts_ops *ops)
144 { 142 {
145 int err = -EEXIST; 143 int err = -EEXIST;
146 struct ts_ops *o; 144 struct ts_ops *o;
147 145
148 if (ops->name == NULL || ops->find == 146 if (ops->name == NULL || ops->find == NULL || ops->init == NULL ||
149 ops->get_pattern == NULL || ops->g 147 ops->get_pattern == NULL || ops->get_pattern_len == NULL)
150 return -EINVAL; 148 return -EINVAL;
151 149
152 spin_lock(&ts_mod_lock); 150 spin_lock(&ts_mod_lock);
153 list_for_each_entry(o, &ts_ops, list) 151 list_for_each_entry(o, &ts_ops, list) {
154 if (!strcmp(ops->name, o->name 152 if (!strcmp(ops->name, o->name))
155 goto errout; 153 goto errout;
156 } 154 }
157 155
158 list_add_tail_rcu(&ops->list, &ts_ops) 156 list_add_tail_rcu(&ops->list, &ts_ops);
159 err = 0; 157 err = 0;
160 errout: 158 errout:
161 spin_unlock(&ts_mod_lock); 159 spin_unlock(&ts_mod_lock);
162 return err; 160 return err;
163 } 161 }
164 EXPORT_SYMBOL(textsearch_register); 162 EXPORT_SYMBOL(textsearch_register);
165 163
166 /** 164 /**
167 * textsearch_unregister - unregister a textse 165 * textsearch_unregister - unregister a textsearch module
168 * @ops: operations lookup table 166 * @ops: operations lookup table
169 * 167 *
170 * This function must be called by textsearch 168 * This function must be called by textsearch modules to announce
171 * their disappearance for examples when the m 169 * their disappearance for examples when the module gets unloaded.
172 * The &ops parameter must be the same as the 170 * The &ops parameter must be the same as the one during the
173 * registration. 171 * registration.
174 * 172 *
175 * Returns 0 on success or -ENOENT if no match 173 * Returns 0 on success or -ENOENT if no matching textsearch
176 * registration was found. 174 * registration was found.
177 */ 175 */
178 int textsearch_unregister(struct ts_ops *ops) 176 int textsearch_unregister(struct ts_ops *ops)
179 { 177 {
180 int err = 0; 178 int err = 0;
181 struct ts_ops *o; 179 struct ts_ops *o;
182 180
183 spin_lock(&ts_mod_lock); 181 spin_lock(&ts_mod_lock);
184 list_for_each_entry(o, &ts_ops, list) 182 list_for_each_entry(o, &ts_ops, list) {
185 if (o == ops) { 183 if (o == ops) {
186 list_del_rcu(&o->list) 184 list_del_rcu(&o->list);
187 goto out; 185 goto out;
188 } 186 }
189 } 187 }
190 188
191 err = -ENOENT; 189 err = -ENOENT;
192 out: 190 out:
193 spin_unlock(&ts_mod_lock); 191 spin_unlock(&ts_mod_lock);
194 return err; 192 return err;
195 } 193 }
196 EXPORT_SYMBOL(textsearch_unregister); 194 EXPORT_SYMBOL(textsearch_unregister);
197 195
198 struct ts_linear_state 196 struct ts_linear_state
199 { 197 {
200 unsigned int len; 198 unsigned int len;
201 const void *data; 199 const void *data;
202 }; 200 };
203 201
204 static unsigned int get_linear_data(unsigned i 202 static unsigned int get_linear_data(unsigned int consumed, const u8 **dst,
205 struct ts_ 203 struct ts_config *conf,
206 struct ts_ 204 struct ts_state *state)
207 { 205 {
208 struct ts_linear_state *st = (struct t 206 struct ts_linear_state *st = (struct ts_linear_state *) state->cb;
209 207
210 if (likely(consumed < st->len)) { 208 if (likely(consumed < st->len)) {
211 *dst = st->data + consumed; 209 *dst = st->data + consumed;
212 return st->len - consumed; 210 return st->len - consumed;
213 } 211 }
214 212
215 return 0; 213 return 0;
216 } 214 }
217 215
218 /** 216 /**
219 * textsearch_find_continuous - search a patte 217 * textsearch_find_continuous - search a pattern in continuous/linear data
220 * @conf: search configuration 218 * @conf: search configuration
221 * @state: search state 219 * @state: search state
222 * @data: data to search in 220 * @data: data to search in
223 * @len: length of data 221 * @len: length of data
224 * 222 *
225 * A simplified version of textsearch_find() f 223 * A simplified version of textsearch_find() for continuous/linear data.
226 * Call textsearch_next() to retrieve subseque 224 * Call textsearch_next() to retrieve subsequent matches.
227 * 225 *
228 * Returns the position of first occurrence of 226 * Returns the position of first occurrence of the pattern or
229 * %UINT_MAX if no occurrence was found. 227 * %UINT_MAX if no occurrence was found.
230 */ !! 228 */
231 unsigned int textsearch_find_continuous(struct 229 unsigned int textsearch_find_continuous(struct ts_config *conf,
232 struct 230 struct ts_state *state,
233 const 231 const void *data, unsigned int len)
234 { 232 {
235 struct ts_linear_state *st = (struct t 233 struct ts_linear_state *st = (struct ts_linear_state *) state->cb;
236 234
237 conf->get_next_block = get_linear_data 235 conf->get_next_block = get_linear_data;
238 st->data = data; 236 st->data = data;
239 st->len = len; 237 st->len = len;
240 238
241 return textsearch_find(conf, state); 239 return textsearch_find(conf, state);
242 } 240 }
243 EXPORT_SYMBOL(textsearch_find_continuous); 241 EXPORT_SYMBOL(textsearch_find_continuous);
244 242
245 /** 243 /**
246 * textsearch_prepare - Prepare a search 244 * textsearch_prepare - Prepare a search
247 * @algo: name of search algorithm 245 * @algo: name of search algorithm
248 * @pattern: pattern data 246 * @pattern: pattern data
249 * @len: length of pattern 247 * @len: length of pattern
250 * @gfp_mask: allocation mask 248 * @gfp_mask: allocation mask
251 * @flags: search flags 249 * @flags: search flags
252 * 250 *
253 * Looks up the search algorithm module and cr 251 * Looks up the search algorithm module and creates a new textsearch
254 * configuration for the specified pattern. 252 * configuration for the specified pattern.
255 * 253 *
256 * Note: The format of the pattern may not be 254 * Note: The format of the pattern may not be compatible between
257 * the various search algorithms. 255 * the various search algorithms.
258 * 256 *
259 * Returns a new textsearch configuration acco 257 * Returns a new textsearch configuration according to the specified
260 * parameters or a ERR_PTR(). If a zero length 258 * parameters or a ERR_PTR(). If a zero length pattern is passed, this
261 * function returns EINVAL. 259 * function returns EINVAL.
262 */ 260 */
263 struct ts_config *textsearch_prepare(const cha 261 struct ts_config *textsearch_prepare(const char *algo, const void *pattern,
264 unsigned 262 unsigned int len, gfp_t gfp_mask, int flags)
265 { 263 {
266 int err = -ENOENT; 264 int err = -ENOENT;
267 struct ts_config *conf; 265 struct ts_config *conf;
268 struct ts_ops *ops; 266 struct ts_ops *ops;
269 267
270 if (len == 0) 268 if (len == 0)
271 return ERR_PTR(-EINVAL); 269 return ERR_PTR(-EINVAL);
272 270
273 ops = lookup_ts_algo(algo); 271 ops = lookup_ts_algo(algo);
274 #ifdef CONFIG_MODULES 272 #ifdef CONFIG_MODULES
275 /* 273 /*
276 * Why not always autoload you may ask 274 * Why not always autoload you may ask. Some users are
277 * in a situation where requesting a m 275 * in a situation where requesting a module may deadlock,
278 * especially when the module is locat 276 * especially when the module is located on a NFS mount.
279 */ 277 */
280 if (ops == NULL && flags & TS_AUTOLOAD 278 if (ops == NULL && flags & TS_AUTOLOAD) {
281 request_module("ts_%s", algo); 279 request_module("ts_%s", algo);
282 ops = lookup_ts_algo(algo); 280 ops = lookup_ts_algo(algo);
283 } 281 }
284 #endif 282 #endif
285 283
286 if (ops == NULL) 284 if (ops == NULL)
287 goto errout; 285 goto errout;
288 286
289 conf = ops->init(pattern, len, gfp_mas 287 conf = ops->init(pattern, len, gfp_mask, flags);
290 if (IS_ERR(conf)) { 288 if (IS_ERR(conf)) {
291 err = PTR_ERR(conf); 289 err = PTR_ERR(conf);
292 goto errout; 290 goto errout;
293 } 291 }
294 292
295 conf->ops = ops; 293 conf->ops = ops;
296 return conf; 294 return conf;
297 295
298 errout: 296 errout:
299 if (ops) 297 if (ops)
300 module_put(ops->owner); 298 module_put(ops->owner);
301 299
302 return ERR_PTR(err); 300 return ERR_PTR(err);
303 } 301 }
304 EXPORT_SYMBOL(textsearch_prepare); 302 EXPORT_SYMBOL(textsearch_prepare);
305 303
306 /** 304 /**
307 * textsearch_destroy - destroy a search confi 305 * textsearch_destroy - destroy a search configuration
308 * @conf: search configuration 306 * @conf: search configuration
309 * 307 *
310 * Releases all references of the configuratio 308 * Releases all references of the configuration and frees
311 * up the memory. 309 * up the memory.
312 */ 310 */
313 void textsearch_destroy(struct ts_config *conf 311 void textsearch_destroy(struct ts_config *conf)
314 { 312 {
315 if (conf->ops) { 313 if (conf->ops) {
316 if (conf->ops->destroy) 314 if (conf->ops->destroy)
317 conf->ops->destroy(con 315 conf->ops->destroy(conf);
318 module_put(conf->ops->owner); 316 module_put(conf->ops->owner);
319 } 317 }
320 318
321 kfree(conf); 319 kfree(conf);
322 } 320 }
323 EXPORT_SYMBOL(textsearch_destroy); 321 EXPORT_SYMBOL(textsearch_destroy);
324 322
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.